The London Underground map is a classic example of a great data visualization.

“Most often, designs that delight us do so not because they were designed to be novel, but because they were designed to be effective; their novelty is a byproduct of effectively revealing some new insight about the world.”

-Noah Iliinsky, Beautiful Visualization

The term ‘data visualization’ definitely isn’t fringe geek-speak anymore — once CNN reports on a trend, it’s pretty much mainstream.

CNN’s feature dates back to 2009; today,  “data visualizations” pop up everywhere — The New York Times, The Guardian, the Los Angeles Times, Twitter, Facebook, Google Maps. An increasing number of enthusiasts, armed with an ever-proliferating arsenal of visual software, are making data come to life.

As data-viz gains momentum (and practitioners), visualizations have taken on fresh, dazzling forms. Colourless scatterplots have been replaced by history flows, heat maps, human figures and hues of every kind.

As an art form, data visualizations will always be in a state of perpetual reinvention. After all, one of the key attributes of a good visualization is making the viewer see data in a new way. But as designers continue to push the boundaries of data perception, the criteria of what makes a great visualization shouldn’t be cast aside.

A few visualization instruments have triggered heated discussions about data-viz criteria in the last few weeks within BuzzData’s local sphere: the first was Microsoft’s Silverlight PivotViewer.

Pivot with purpose

Earlier this month, Microsoft Canada open-source strategy lead Nik Garkusha (and team member of Open Halton) introduced a visualization to Open Hamilton’s discussion group that put Vancouver’s council expense data through PivotViewer: “I wanted to learn how to use PivotViewer and to build dynamic CXML collections, which was a fun learning experience,” he said, asking for feedback:

(See Garkusha’s visualization live here.)

Garkusha said he was originally motivated by Gary Flake’s TED talk demo of PivotViewer from 2010:

Flake’s TED talk certainly gives the impression that PivotViewer’s animation and extreme-zoom features are capable of making almost any data set pop. However, the group’s feedback was a good reminder that effective data visualization is less about sophisticated software and more about carefully considering the data you have, and what knowledge you want people to glean from it.

“It’s certainly pretty, but it’s not clear to me why this is a good visualization,” open-data hacker James McKinney said. “How do I answer simple questions like: Which councillor spent the most money, or who spends the most on parking?

“What do I care how frequently a councillor files expenses?” McKinney asked, referring to PivotViewer’s predominant use of the COUNT function over than the SUM function. “‘Who spent more money?’ is a far more common and important question than ‘Who spent the most frequently?’ “

Before long other group members, including Garkusha, began debating the issue, deconstructing the merits and faults of putting expense data in PivotViewer format.

Garkusha countered that while how much a councillor spends is certainly a common question, a visualization’s value is subject to what each viewer is looking for, in which context: “I do believe that like art, visualizations are interpreted and understood differently by different people, and whether it’s ‘good’ or ‘bad’ depends on the questions they are looking to answer & insights they are looking to derive.

“If the objective is to answer ‘which councillor spent the most money,’ what I put together as a visualization is ‘bad.’

“If the objective is to answer ‘which councillor submitted most expenses,’ what I put together may not be so bad.

“I do think it’s a bit shortsighted to imply that a certain type of question is more relevant or important than others,” he added. “Who decides what’s relevant?”

Personally speaking, I think both McKinney and Garkusha were homing in on one central point about creating data visualizations — the most telling criterion of a “good” or “bad” viz is whether or not it’s apparent the creator considered what is relevant (or worth learning) in their data, for whom, and of course, which design or tool is the right way to communicate it.

In the PivotViewer video, for example, Flake shows spectacular visualizations out of screenshots of browsing history — but browsing history is very visual and frequency-oriented, i.e.: “how often do you check which pages?” Naturally, PivotViewer’s COUNT function and deep-zoom would be perfect for this kind of data.

City council expense data, on the other hand, might need nothing more than a few clean tables, as McKinney pointed out at one point in the discussion. “You
don’t need to make data fun to make it interesting,” he added.

-Momoko Price